APPLICATIONS OF LEVEL SET METHODS FOR IMAGE AND VIDEO PROCESSING

In this paper, the level set algorithm is described along with its application in areas of image and video processing. The level set algorithm is a form of active modelling, such as active contours or ‘snakes’. The snake can be viewed as Lagrangian geometric formulation where the boundary of the model is defined in parametric form. In the level set algorithm, this Lagrangian formulation is replaced with an Eulerian, initial value partial differential equation, which is more robust, than snakes, in tackling certain image processing applications. This paper describes extension to the theory of algorithm application to feature extraction and object tracking. In applications where off-line computation is acceptable, it has been shown that the level set algorithm competes well with established techniques, but can be demonstrated to have many advantages over them. A disadvantage of the level set method over active contours has been found to be the high computational cost.